Latency and Relative Frequency of Response in Psychophysical Discrimination

Abstract

The relation between response latency and probability is discussed with particular reference to differences between mean times for correct and incorrect responses in difficult two-choice discriminations. The predictions of some finite Markov chain models for this situation are considered, and possible modifications of these simple systems which may bring about features of observed data are discussed. Some results, in the form of curves of the latency-probability relation, are presented from earlier work and new discrimination experiments. These data are discussed in relation to theoretical interpretations based on a particular stochastic model which describes psychophysical discrimination in terms of the accumulation of units contributing to the overt response.

The latencies of correct and incorrect responses in discrimination and detection tasks: Their interpretation in terms of a model based on simple counting

“…Atkinson et Audley & Pike, 1965;Laberge, 1962) will be examined for its ability to account for latency features in difficult two-choice discrimination, and a preliminary comparison will be made with Experiment 1 was originally performed to assess the validity of different kinds of stochastic model for discrimination behavior (Pike, 1967). The accumulator model appeared to fit the data best, and the results from other work (e.g., Kellogg, 1931) also supported the model.…”

“…It should be apparent from the graphs that good evidence exists for the operation of a simple counting process in difficult discrimination, certainly so far as comparison with other more complicated stochastic processes is concerned (cf. Pike, 1968). It should be pointed out, however, that the individual graphs arc not, on the whole, as clearly of the same form as illustrated here, and two of the seven Ss have graphs intermediate in form between Type A and Type B.…”

“…One prominent measure of latency data in choice tasks is the quantitative difference between correct and incorrect responses. This tends to differ with particular tasks: in choice reaction time, it appears that correct response latencies are, on the average, longer than incorrect latencies, and attempts have been made to usc or explain this fact by theoretical arguments (e.g., Laming, 1968); in psychophysical discrimination (e.g., the judgment of which of two stimulus intensities is the greater) the opposite appears to be the case, so that correct mean latencies are shorter, at least when the discrimination is difficult (Kellogg, 1931;Pike, 1968); with detection tasks the matter may depend upon whether S + N or N is presented (Carterette et al, 1965). In this paper a particular stochastic model of the "random walk" type (cf.…”

“…The above description in terms of discrete time intervals may be modified to account for a process in which events a and b occur in nondiscrete or random fashion (Pike, 1968). However, the results concerning comparative mean latencies and response probabilities remain the same, and hence, the LP graphs are not altered.…”

“…The accumulator model appeared to fit the data best, and the results from other work (e.g., Kellogg, 1931) also supported the model. Experiment 2 was subsequently performed to investigate the situation where Ss were uncertain as to whether speed or accuracy of response was required, and curve from P =0 to P =1 is a particularly important feature of the model, since other random walk models predict that for low values of P (i.e., incorrect responses associated with less difficult discrimination) the latencies are faster than for P = 0.5 (Pike, 1968). A model derived from signal detection theory which also predicts the Type A LP curve will be discussed later.…”